Replit CEO details path to artificial developer intelligence, raises new $20M investment

Developer platform company Replit announced today that it has raised $20 million from Craft Ventures. The new investment comes as Replit continues to advance its agenda to enable developers with generative AI capabilities as it builds towards a future of artificial developer intelligence (ADI).

The $20 million investment is not a typical funding round for a startup. In fact, the new investment isn’t about raising new money, but rather is a liquidity event for some of the company’s longer term employees. Replit was founded in 2016 and its last major funding round was in April when the company raised $97 million, which gave the company a valuation of $1.16 billion. In many early stage startups, the earliest employees are offered some form of equity or shares. Typically the only way those shares become ‘liquid’ or cashable is if the company is acquired or has an initial public offering (IPO). The new $20 million is an opportunity of equity holding Replit employees to cash out if they so choose. 

The new investment comes several weeks after Replit announced its “AI for All” initiative which integrated the company’s developer AI capabilities for all of its users. Replit develops its own large language model (LLM) known as replit-code that helps with code generation. The company is now gearing up to detail new efforts to help enable developers to become even more productive, with the power of AI to back them up.

“One thing we try to make clear to the world, to ourselves and to our employees is that we’re not in the business of selling AI, OpenAI, Anthropic and those sorts of companies can focus on that,” Amjad Masad, CEO of Replit, told VentureBeat. “We are in the business of selling a dream, which is to make your dream software more accessible and make programming more accessible.”

Why Replit built its own LLM for code

Replit’s coding LLM technology competes against multiple rival technologies such as Github Copilot, Amazon CodeWhisperer and open source efforts like StarCoder and the Code Llama project.

Masad emphasized that the Replit approach is somewhat differentiated in that his company has a whole platform. Developers not only use Replit to write code, they also can use it to deploy and run code in production. 

Replit also aims to differentiate by building models tailored to its own platform usage data, which includes unique runtime information not available elsewhere. Replit’s own data from usage on its platform allows it to fine-tune models specifically for the types of tasks and code seen on Replit. Masad noted that this gives Replit an advantage in building a superior product because of the data that is on the platform.

The 1000x developer and why learning to code is still important

A common question that has come up in the past year in the market as a whole, is what the role of developers will be moving forward, in a world where code can be automatically generated.

While AI promises significant boosts in productivity, Masad does not expect the need for developers to diminish any time soon, if ever. In fact, he’s a proponent of the concept known as the 1000x developer, that is a developer that can be significantly more productive, thanks to the power of AI.

“My view is that learning to code actually has a better return on investment right now than it had a year ago,” Masad said. “The reason is because you learn a little bit of coding, and then you get this massive boost from AI.”

Previously, he noted that an individual had to learn a lot in order to get to a level where it was possible to make an application. Masad said that he has seen people that start learning how to code on Replit and are quickly building applications and even full scale businesses.  

Though AI-powered code generation tools are powerful, Masad also emphasized that today there is still a clear need for a human in the loop to build applications and developers will continue to benefit from learning how to code for many reasons.

“There’s always going to be edge cases where we’re going to have to drop into the code,” he said. “And so learning how to code is going to be essential to understand what’s happening behind the scenes to catch those cases.”

Agents and the path to Artificial Developer Intelligence (ADI)

Replit is now gearing up for its next Developer Day event on Nov. 14 where the company will detail progress to date and outline a vision for the future.

One of the things that Replit will be talking about is the company’s approach toward what are commonly referred to as AI agents, which are becoming increasingly popular ways of automating different tasks and extending AI.

That agent approach could involve giving access to an LLM to tools that humans use so they can install software packages, manage application runtimes and automatically deploy code. At Replit, the company calls its approach to AI agents – Artificial Developer Intelligence.

“There are a lot of companies focused on AGI [artificial general intelligence], we think what’s more within reach and definitely on brand for us, is the vision of artificial developer intelligence,” Masad said. “It’s about really creating a bunch of helpers, a bunch of co workers that could work with individual engineers that are actually AI agents.”